The predictive value of the preoperative C-reactive protein-albumin ratio for early recurrence and chemotherapy benefit in patients with gastric cancer after radical gastrectomy: using randomized phase III trial data

Background The definition and predictors of early recurrence (ER) for gastric cancer (GC) patients after radical gastrectomy are unclear. Methods A minimum-p value approach was used to evaluate the optimal cutoff value of recurrence-free survival to determine ER and late recurrence (LR). Receiver operating characteristic curves were generated for inflammatory indices. Potential risk factors for ER were assessed with a Cox regression model. A decision curve analysis was performed to evaluate the clinical utility. Results A total of 401 patients recruited in a clinical trial (NCT02327481) from January 2015 to April 2016 were included in this study. The optimal length of recurrence-free survival to distinguish between ER (n = 44) and LR (n = 52) was 12 months. Factors associated with ER included a preoperative C-reactive protein-albumin ratio (CAR) >= 0.131, stage III and postoperative adjuvant chemotherapy (PAC) > 3 cycles. The risk model consisting of both the CAR and TNM stage had a higher predictive ability and better clinical utility than TNM stage alone. Further stratification analysis of the stage III patients found that for the patients with a CAR < 0.131, both PAC with 1-3 cycles (p = 0.029) and > 3 cycles (p < 0.001) could reduce the risk of ER. However, for patients with a CAR >= 0.131, a benefit was observed only if they received PAC > 3 cycles (54.2% vs 16.0%, p = 0.004), rather than 1-3 cycles (58.3% vs 54.2%, p = 0.824). Conclusions A recurrence-free interval of 12 months was found to be the optimal threshold for differentiating between ER and LR. Preoperative CAR was a promising predictor of ER and PAC response. PAC with 1-3 cycles may not exert a protective effect against ER for stage III GC patients with CAR >= 0.131.